Perhaps kieranmillar can provide some more insight on this, since he has created a bunch of L2 levels?
Hell yeah. I'm someone who has opinions.
Before I begin I should make 2 things clear: 1) In my level pack I deliberately imposed onto myself stringent limits on what skills could be used, so often had no luxury in choosing between the platformer and builder. 2) L2 has significant limits on what you can do with terrain due to its grid system, while in L1 you have pixel precision in your layouts.
Height gaining with platformsIn L2 where I have less control over terrain, platform stacking is horrible. It's at its worst when trying to do interesting things with skills like the magno booter where tiny thin gaps really matter a whole lot, and has other unfortunate consequences for skills like the pole vaulter or skier where tiny height changes can really matter, etc. It's true that if the platformer were completely flush with the ground you wouldn't be able to platform over traps, but honestly there are so many other skills in L2 that can fulfil that task, like the filler, that it doesn't really matter in the big picture. You'd easily live without it.
I really like that NeoLemmix's platformer can't be used to gain height, it's definitely its biggest strength. It ensures that the platformer is actually meaningfully different from the builder in an important way. The builder is extremely powerful in how multi-purpose it is, the platformer is a lot less powerful but often times being less general-purpose is much more useful for a designer. It's certainly less prone to endless backrouting.
The fact that L2's platformer doesn't have to be used right at the edge of a platform is also handy though, especially given that L2 doesn't have any replay or rewind functionality at all. If it was like Neolemmix then you'd be constantly missing your assignment and be forced to redo the whole level.
In this regard, if I were designing Neolemmix's platformer from scratch I'd make one key change in this regard, I'd allow a platformer to build even if the brick he was placing was placed entirely inside terrain. This would let you start building from anywhere on flat terrain, and when you think about it, is consistent with how the builder works too, where what triggers the lemming to cancel building is not that the brick cannot be placed, but the lemming himself is blocked from moving by terrain. This would also mean the lemming would not stop platforming if he reached the other side of a gap at the same height and so not turn around, but that's fine too I think, if it had been this way from the start I don't think anybody would have complained.
Turning around when hitting terrainWhen designing my L2 levels, I wanted to focus on interactions between the skills for two main reasons, 1) the skills are like the whole point of the game and 2) I wanted to focus on the ways in which L2 was different from L1, and the skills were the primary difference.
Often I was faced with an interesting design conundrum, in order to utilise the interaction between two skills, I would first need one lemming to use one skill, then have another lemming come along and finish the job with another skill, but couldn't have the first lemming be the one to finish the job or else the interaction wouldn't work or be interesting. But often this meant somehow ensuring that the first lemming behaved differently to the second lemming in a fundamental way. My most common solution was to make the first lemming a climber, so that he is forced to take a different path by not being able to turn around as easily. I use this technique quite a lot. And now the magic of QFK2 is ruined as you notice many levels are just formulaic differences on the same basic design, I just spoiled it all sorry
And one thing I discovered early on is that the fact that the platformer does not turn the lemming around when stopping due to hitting terrain was incredibly useful for enforcing this. In the first race where geoo and Simon played the Sports tribe, it was funny (or maybe depressing?) to watch them spend 40 minutes on Pole Position, a level where I had to use the builder, trying to turn around a climber instead of doing the actual cool and interesting solution. Oh the horrors that could have been avoided if the builder didn't let you turn a lemming around. It turns out that this difference is actually a lot more meaningful than you might at first think. By not having to worry about arbitrarily turning around a lemming, you can make the puzzle more focussed, cleaner, and more forgiving, instead of every little bump in terrain being a backroute hazard.
A really important lesson that the design of L2 teaches us (by getting it so horribly wrong) is this: Making skills is all about maximising utility while minimising the actual number of different pieces. L2 has a huge number of skills that are often barely different from each other in all but a few nitpicky differences that it actually ends up a bit worse off as a result than if it didn't have them. Part of this is because it is almost inevitable when making puzzles for video games to focus on the nitpicky mechanical differences, I too am guilty of this, so the more the differences can be made clear by having less parts overall, the cleaner things will be for everyone and the more time will be spent on thinking about the logic of the puzzle and not mechanical nuance. Meanwhile L1's skillset is supremely focussed and well-rounded, and while there is welcome room for a bit of expansion, you want to ensure maximal impact of each skill so that you can cover your bases with fewer moving parts. This is especially true if the new skill covers some similar ground (in this case, being another skill for creating terrain), you want the difference between the two skills to be as meaningful as possible so you don't have to add even more skills to fill any mechanical gaps.
If I were designing Neolemmix's platformer from scratch, I'd change it in this regard by not having the lemming turn around when it is forced to stop building. That way if all your level needs a construction skill for is to cover a gap, you sometimes have the option to add one without also running the risk of neutering the single powerful mobility skill, the climber, letting players focus on the fundamentals of your puzzle and things being overall a bit cleaner.
Incidentally, the fact that L2 miners don't turn when hitting steel was also the preferred behaviour for me almost all the time (no I'm of course not advocating for any changes here). If I want the lemming to turn, I have general terrain for that. Similarly with Neolemmix having walker and cloner skills, how many more turning methods do you really need?